Overview

This is the analysis of white matter using FSLs TBSS


Methods

Participants

  • Total number of scans = 120 (males and females)

  • Total scans used in analysis= 118

  • Excluded subs 329 and 372 from analyses due to bad behavior data and Subject 347 failed quality control check in DSI


Scan parameters

  • Locations from UCSB and UCI

QSIprep

  • Version used: 0.16.0RC3

  • The following parameters were used

    • dwi_denoise_window 5
    • hmc_model 3dSHORE
    • hmc-transform Rigid
    • shoreline_iters 2

Preprocessing for FSL

The data needs to be reconstructed and prepared for FSL.

  • Run FSL dti fit in order to extract our statistics. It fits a tensor model at each voxel.
    • You need the following inputs
      • DWI scan images
      • BRain mask image
      • Output name
      • Bvec files
      • B value files
  • We need to reorient the images to be in the same format as FSL atlases.
    • We run FSL reorient on each image


FSL TBSS

  • Step 1: tbss _1_preproc *nii.gz
    • this prepares the data for full preprocessing

  • Step 2: tbss_2_reg
    • This runs a nonlinear registration and creates transforms of all the images in order to put them into a standard space

  • Step 3: tbss_3_postreg
    • Applies all of the transofrms created in the previous step

  • Intermediate Step: Quality check
    • Open all images in FSL and visually inspect that they line up with the mean FA

  • Step 4: tbss_4_prestats
    • Creates a threshold for each of the masks in order to extract stats

  • Custom scripts
    • here we use custoim fsl maths and meants scripts in order to extract our values for the JHU atlas



Results

Overall approach

FA

First, Let me plot and check all of our FA data

histograms for ROIs

histograms for ROIs split by sex

Scatter plots against age for ROIs

Maze Accuracy

Let’s look at the distribution for maze accuracy behavior. Data appears to be not normally distributed.

## Warning: Removed 36 rows containing non-finite values (`stat_bin()`).

## Warning: Removed 36 rows containing non-finite values (`stat_bin()`).

Let’s just double check that by running a normality test.

The data is not normally distributed with p = 6.1641918^{-13}.

Checking independency

Checking normality

ROI p.value
Middle_cerebellar_peduncle 0.0090485
Splenium_of_corpus_callosum 0.0018326
Fornix_(column_and_body_of_fornix) 0.0378160
Inferior_cerebellar_peduncle_R 0.0165339
External_capsule_R 0.0076241
External_capsule_L 0.0489276

partial correlations

Running semi correlations using ppcor sex and age are covariates
ROI estimate p.value statistic n gp Method
Middle_cerebellar_peduncle -0.2484085 0.0262958 -2.264876 82 2 pearson
Medial_lemniscus_R -0.2246881 0.0450978 -2.036462 82 2 pearson
Superior_fronto-occipital_fasciculus_(could_be_a_part_of_anterior_internal_capsule)_R -0.2700141 0.0154250 -2.476693 82 2 pearson
Superior_fronto-occipital_fasciculus_(could_be_a_part_of_anterior_internal_capsule)_L -0.2658221 0.0171622 -2.435294 82 2 pearson

Mediation analysis

We want to know whether sex mediates the relationship between FA and Maze Accuracy in midlife adults. in this case X= ACC, Y = FA and M = sex. Looks like Sex does not mediate the effect of FA on ACC in middle cerebellar peduncle.

  1. Let’s run a linear model using FA and Acc.

DSP

partial correlations

Running semi correlations using ppcor sex and age are covariates
ROI estimate p.value statistic n gp Method

MD